Experiments using MMX-based processors for real-time image processing on the ARGO vehicle

This paper presents the preliminary results of a study concerning the impact of MMX technology on mobile computing for intelligent vehicles applications, which, because of their hard real-time constraints, is an undoubtly challenging task. More precisely, the low level portion of the processing aimed at the detection of obstacles in the GOLD system, which is the software system that manages the activities of the ARGO autonomous vehicle, has been implemented exploiting the MMX features. A comparison with a traditional scalar code is discussed and the advantage of MMX-based solution presented along with some examples. Keywords| MMX, autonomous vehicles, machine vision, Inverse Perspective Mapping, stereo vision. I. Introduction C URRENT general-purpose processors have been enhanced with new features explicitly dedicated to the eÆcient handling of multimedia (audio and video) data: in order to exploit the high intrinsic spatial parallelism of low level image processing, a SIMD solution is uxed.